Math Reasoning in LLMs is Organized by Approach, Not Topic
Mathematical reasoning benchmarks are typically organized by topic, but language models may organize their internal computation by reusable reasoning approach instead. In this paper, we investigate whether open math-capable LLMs organize internally by topical sub-skill or by reasoning approach, and we present evidence...
Sajad Goudarzi, Samaneh Zamanifard, Moloud Nasiri et al.
· 0 citations